Predictive Modeling of Air Quality: Performance Comparison of Supervised Machine Learning, Deep Learning, and Extreme Learning Machines

This research study presents a comprehensive analysis comparing the performance of supervised machine learning, deep learning, Extreme Learning Machines (ELM), and Genetic Algorithm (GA) models in predicting the Air Quality Index (AQI). With growing concerns about air pollution and its severe impact on both public health and the environment, developing accurate and timely forecasting models has become critical for effective intervention strategies. This study leverages real-world datasets to assess the predictive accuracy, computational efficiency, and applicability of these different models in various forecasting scenarios. We focus on key attributes such as sulfur dioxide (SO2), nitrogen dioxide (NO2), respirable suspended particulate matter (RSPM), suspended particulate matter (SPM), and PM2.5 to predict AQI. The results indicate that deep learning models excel in predictive accuracy, while ELM models offer an optimal balance between accuracy and computational speed, making them suitable for real-time applications. Genetic Algorithms, known for their optimization capabilities, also demonstrate promise in refining model parameters, further improving the forecasting performance. Additionally, supervised machine learning techniques, including decision trees, support vector machines, logistic regression, random forest, naïve Bayes, and k-nearest neighbors, continue to show reliable results, underscoring their importance in environmental forecasting tasks. These findings provide a comparative perspective, aiding in the selection of appropriate models for air quality prediction depending on specific requirements for accuracy, efficiency, and scalability.